Modelling and evaluation of mechanical performance and environmental impacts of sustainable concretes using a multi-objective optimization based innovative interpretable artificial intelligence method

Abstract

This study focuses on modelling sustainable concretes' mechanical and environmental properties with interpretable artificial intelligence-based automated rule extraction, management of waste materials, and meeting future prospects. In this context, 24 sustainable concrete series containing fly ash and recycled aggregates were produced. Compressive strength tests were performed on these specimens at 7, 28, and 90 days, and their mechanical properties were evaluated. Concrete classes (Class A, B, C, D) were determined using the compressive strength values obtained for each test day. The results of each concrete class were analyzed using a unique interpretable multi-objective rule extraction model, and the range values of the materials used were determined. The applied multi-objective rule extraction method is used for the first time in the literature, and its most important novelty is that, unlike other black-box artificial intelligence methods, it can also enable the creation of sustainable concrete recipes. After the range values of the materials used were found by automatic rule extraction, environmental impact assessments were performed. Among the impact categories, energy consumption and global warming potential were considered. The energy consumption results for Rule 4 were calculated as 814.8–1467.1 MJ, respectively, and a reduction of approximately 44.5% was observed. Similarly, global warming potentials for Rule 3 were obtained as 187.0–267.3 kg m³, respectively, with a reduction of about 30%. The limitations and future prospects of the study have been extensively investigated. The importance of adopting explainable/interpretable artificial intelligence-based approaches within the scope of sustainable development and circular economy goals to develop social infrastructure and buildings with low carbon emissions that are feasible in terms of mechanical and environmental properties is highlighted. Multi-Objective Optimization Based Innovative Interpretable Artificial Intelligence Method, used for the first time in mechanical and environmental modelling of sustainable concretes, can make significant contributions to the literature and future studies.